CRITIQUE OF THESIS

“Transaction Cost Theory and Just-in-Time

Manufacturing: A New Look at Vertical
Integration in the United States Automobile
Market” by T.H. Klier, Michigan State
University, 1993

A critique submitted in partial fulfilment

of the requirements for the degree of
Doctor of Business Administration

by

Staffan Canbäck

Henley Management College

Brunel University

March 1997

© 1997 Staffan Canbäck

 2

CONTENTS

1. INTRODUCTION ...................................................................................... 3

2. EXECUTIVE SUMMARY.......................................................................... 4

3. REVIEW OF PROBLEM AND HYPOTHESIS DEFINITION ........... 6

3.1 IMPORTANCE OF PROBLEM ......................................................... 6
3.2 CLARITY OF PROBLEM AND HYPOTHESIS STATEMENTS .. 8
3.3 LINK TO THEORY ............................................................................. 9
3.4 RESEARCHABILITY ........................................................................ 10

4. LITERATURE REVIEW .......................................................................... 12

4.1 THE PARADIGM SHIFT: FROM FORDISM TO JUST-IN-TIME12
4.1.1 Fordism .................................................................................. 13
4.1.2 Just-in-Time Manufacturing................................................ 14
4.1.3 Implication of Changes ........................................................ 15
4.2 TRANSACTION COST THEORY .................................................. 17
4.2.1 Theory .................................................................................... 17
4.2.2 Empirical Studies .................................................................. 18

5. ANALYTICAL APPROACH .................................................................. 21

5.1 ANALYTICAL METHOD ............................................................... 21
5.2 DATA COLLECTION ...................................................................... 21
5.2.1 Current Data .......................................................................... 22
5.2.2 Earlier Studies ....................................................................... 22
5.3 STATISTICAL ANALYSES ............................................................. 23
5.3.1 Comparison to Previous Studies ........................................ 24
5.3.2 Just-in-Time’s Impact on Governance ............................... 26
5.4 MEASUREMENT OF VARIABLES................................................ 28
5.5 QUALITY OF STATISTICAL ANALYSIS..................................... 29

6. FINDINGS ................................................................................................. 31

7. THESIS LAYOUT ..................................................................................... 34

8. CONCLUSION ......................................................................................... 36

REFERENCES ..................................................................................................... 37
 3

1. INTRODUCTION

The following is a critique of Thomas Helmut Klier’s PhD thesis

“Transaction Cost Theory and Just-in-Time Manufacturing: A New Look

at Vertical Integration in the United States Automobile Market”, Michigan

State University (Klier 1993). I have chosen to critique this thesis for three

reasons. First, my own research investigates the possible limits to size in

corporations by applying transaction cost theory to explain size

disadvantages. Klier’s thesis builds on the same theoretical foundation

and his review of the transaction cost literature is highly relevant to my

own research effort. Second, Klier operationalises transaction costs in a

similar way to what I expect to do in my research. I thus hope to build on

his methodology when I design my test of the hypotheses in my research.

Third, Klier discusses his statistical approach thoroughly and I expect to

use a similar approach in my research. In particular, he uses both a

binomial and an ordered probit model that I anticipate will be of use in my

research.
 4

2. EXECUTIVE SUMMARY

Klier has written a concise, well-structured, and thoughtful thesis that

enhances our understanding of vertical integration and the role

transaction costs play in make-or-buy decisions. It is an important

contribution to knowledge since it is the first explicit analysis of how

vertical integration is affected by just-in-time techniques in manufacturing.

Klier shows that just-in-time techniques necessitate increased mutual

commitment between OEMs and suppliers. This increase in commitment

leads to quasi-integration1 in the vertical chain, at the expense of market-

based arms-length relationships characteristic of the Fordist era.

Furthermore, one of the key problems in the field of transaction cost

theory is that there have only been a few attempts at operationalising the

theory’s concepts, and there are even fewer empirical studies. Klier’s

thesis adds to the knowledge in these areas and can usefully serve as a

model for similar future studies.

The thesis is thoroughly researched and builds on an extensive survey of

literature and a limited, but specific, empirical analysis based on a

qualitative questionnaire sent to automotive manufacturers. The

1 Quasi-integration is defined as long-term relationships between fairly autonomous but mutually

dependent partners. This is distinct from vertical integration where the production process is
fully internalised, and "no integration" where upstream and downstream firms have a purely
contractual, market-based, relationship (cf. Klier 1993, 3-4).
 5

hypothesis is tested creatively against earlier research and the new

questionnaire data.

The statistical methods employed have been meticulously chosen and

applied. However, the fairly low response rate to the questionnaire and

the consequently small sample size, make it difficult to draw

unambiguous conclusions. Seen as a whole, though, the thesis adds

insights to the understanding of transaction costs and provides inspiration

for further research.

 6

3. REVIEW OF PROBLEM AND HYPOTHESIS

DEFINITION

Klier identifies a crucial problem that has not been researched before, and

articulates a testable hypothesis. The problem formulation is in line with

earlier research and builds on the traditions of, for example, Monteverde

and Teece (1982), and Masten, Meehan and Snyder (1989) who have

studied related, but distinctly different, problems.

3.1 IMPORTANCE OF PROBLEM

The introduction of just-in-time techniques in the United States

automotive industry has fundamentally changed manufacturing and

sourcing arrangements. One survey (Helper 1991) shows that between

1984 and 1989, the industry was transformed, as exemplified by the

increased information exchange between OEMs and suppliers, the shift in

primary purchasing criterion from price to quality, and the lengthening of

contract times. A key outcome of this is that vertical integration has

decreased, while supplier relationships reputedly have become more co-

operative in nature: “these results indicate progress toward a voice model2

of supplier relations, in which suppliers play an important role in solving

joint problems and in generating fresh ideas about products and

2 As defined by Hirschman (1970).

 7

processes.” However, little research has been done on the linkage between

the introduction of just-in-time techniques and the changing buyer–

supplier relationships, and most information is anecdotal.

Vertical integration problems are also fertile ground for testing transaction

cost theory. The theory predicts that vertical integration, rather than the

use of market transactions, will be more common when the assets of the

buyer and supplier are linked and interdependent. This so-called asset

specificity can be in the form of site,3 physical asset,4 human asset5 and

dedicated asset specificity6 (Williamson 1985, 95–96). With the increased

use of just-in-time techniques, it is expected that the asset specificity

increases and thus new co-operative forms of relationships between

buyers and suppliers evolve, being closer to vertical integration than arms-

length, market-based relationships.

It is argued that the new manufacturing system is

characterized by a high degree of mutual commitment
between up- and downstream firms, leading to the formation
of market-based vertical relationships. (Klier 1993, ii)

3 Site specificity arises when successive stages of production need to be located in close proximity
to each other and the set-up or relocation costs are great.
4 Physical asset specificity is high if assets are dedicated to a specific purpose and can not easily be
used in other operations.
5 Human asset specificity exists if human resources are trained specifically for a stage in the
production process and the skill cannot be used elsewhere.
6 Dedicated asset specificity exists when a supplier expands capacity based on demand from a
particular buyer. Klier does not refer to dedicated asset specificity.
 8

The results support the hypothesis that that the arrival of

just-in-time manufacturing influences the decision to
vertically integrate. The degree of mutual commitment is
consistently found to be a significant determinant of
governance structure. The presence of high degrees of
mutual commitment, typical for just-in-time manufacturing,
strengthens the ability to enforce contractual agreements by
making hold-up threats less credible. In doing so, it increases
the self-enforcing range of contracts. (Klier 1993, iii)

Consequently, understanding the buyer–supplier relationships in the

automotive industry from a transaction cost perspective can be an

excellent test of the theory’s predictive power. It can also contribute to

making transaction cost theory more operational, a challenge the theory’s

proponents have struggled somewhat unsuccessfully with over the last

twenty years.

Klier thus addresses an important issue and his research passes any test of

importance and contribution to knowledge.

3.2 CLARITY OF PROBLEM AND HYPOTHESIS

STATEMENTS

The problem stated by Klier—has the introduction of just-in-time

techniques changed the nature of vertical relationships in the US

 9

automotive industry?—is clearly expressed and succinct. He converts this

problem into a hypothesis:7

...the introduction of JIT manufacturing can effectively

reduce the propensity for opportunistic behavior by
increasing the extent of mutual commitment present in
vertical relationships. Therefore I expect an increase in
mutual commitment to result in a decrease in the probability
of vertical integration. (Klier 1993, 65)

Both the problem and the hypothesis are delimited and narrow since they

address only one aspect of buyer–supplier relationships in a specific

industry (automotive) and geography (United States). At the same time

they are exhaustive and researchable. It is difficult to imagine other

problems or hypotheses that should have been tested at the same time.

3.3 LINK TO THEORY

The problem and hypothesis are well linked to the theoretical foundation.

Klier builds on Williamson’s (1985) framework for understanding vertical

relationships and complements this with four other theoretical sources to

argue the expected outcome of the research.

7 Klier's hypothesis can be misunderstood since a decrease in vertical integration might lead the
reader to believe that there will be a commensurate increase in market-based relationships (no
integration). However, the proportion of such relationships will also decline according to Klier's
hypothesis. The difference is made up of quasi-integration.
 10

The key negative criticism is that Klier accepts Williamson’s framework

without testing its validity, or even recognising that the validity can be

challenged. An example is Williamson’s argument that increased

uncertainty leads to increased vertical integration, which could be

questioned since we have seen increased turbulence in the automotive

market over the last 20 years, while vertical integration has decreased.

Balakrishnan and Wernerfeldt (1986) demonstrated, without taking a

transaction cost perspective, that technological uncertainty leads to less

vertical integration. Walker and Weber (1984) made the same observation

in empirical studies of make-or-buy decisions in the automotive industry.

Mahoney (1989, 1992), in reviews of the transaction cost literature on

vertical integration, argues that high uncertainty leads to less investment

in specialised assets and consequently lower vertical integration. Clearly,

Williamson’s view can be challenged.8

3.4 RESEARCHABILITY

The problem is researchable since Klier breaks down his hypothesis into

five sub-hypotheses, each of which can be tested quantitatively and

without massive amounts of research. Furthermore, these sub-hypotheses

8 A useful distinction may be to divide uncertainty into at least two components: volatility
uncertainty and variety uncertainty. Volatility refers to temporal fluctuations, variety to degrees
of freedom in technological and market choices.
 11

are aligned with earlier research so that older data can be used to support

or dismiss the findings.

Klier also recognises the limitations of his research. For example, he states

that the changes in the manufacturing paradigm and in the determinants

of governance structure may be driven by an underlying common source,

but that he is not addressing this issue.

 12

4. LITERATURE REVIEW

The literature survey is thorough and to the point. It covers the relevant

literature of just-in-time manufacturing and transaction cost theory.

However, it does not recognise alternative theoretical approaches such as

agency theory and technology-driven explanations of vertical integration.

Klier chooses to divide the literature survey in two parts. First he discusses

the general changes in the manufacturing paradigm in the United States,

showing how the just-in-time philosophy overtook traditional Fordist

manufacturing during the 1980s. He then discusses transaction cost theory

and how the paradigm shift should affect vertical integration if transaction

cost theory holds true.

4.1 THE PARADIGM SHIFT: FROM FORDISM TO

JUST-IN-TIME

Chapter 1 of the thesis discusses the manufacturing paradigm shift. It

focuses on the key aspects of relevance to the research and leaves out

general discussions of the changes in the US automotive sector.9 This

makes the overview comprehensive, yet targeted. It achieves its purpose

of positioning the issues at hand.

9 Klier has a footnote on general discussions on page 12.

 13

4.1.1 Fordism

Klier summarises the key aspects of the Fordist system on pages 13–17. He

argues that Fordism is built on two principles: the separation of

intellectual and manual work, and the specialisation of labour. The

application of these principles resulted in the assembly line approach to

manufacturing. The implication of the assembly line is that manufacturers

favour stable production volumes and buffer inventory to avoid

disruptions. As a consequence, General Motors, Ford, and Chrysler tended

to 1) be vertically integrated, 2) have an arms-length relationship with

suppliers, and 3) negotiate short-term (1 year) contracts with their

suppliers.

A number of sources are used to illustrate these points. Most importantly,

Cohen and Zysman (1987) are referenced on the buffer tendency, Scherrer

(1991) is used to prove the vertical integration point, and Womack, Daniel

and Roos (1990) and the MIT Commission (1989) are used as evidence for

the supplier relations points. Together, these references suffice to support

Klier’s view on how the Fordist system works and this section of the

literature review is appropriate in length and scope. The only issue may be

that the definition of vertical integration used by Klier (value added 

revenue) is somewhat discredited (Caves and Bradford 1988), especially

when comparisons are made between industries.

 14

4.1.2 Just-in-Time Manufacturing

Klier uses pages 17–25 to characterise just-in-time manufacturing and

contrast it with the Fordist system. His conclusion is that the new system

represents a paradigm shift and that almost all attributes of the

manufacturing approach have changed.

While Fordism built on separation and specialisation, just-in-time

manufacturing is based on flexibility to changes in demand, according to

Klier. From this stems a number of practices that differ sharply from the

Fordist system. Teamwork is at the core of the system and the separation

of intellectual and manual labour is less pronounced. Inventory is seen as

a problem rather than a solution. Subcontractors are used for a large

proportion of value added. To optimise the use of subcontractors, the just-

in-time driven automotive manufacturer tends to have a closer

relationship with its suppliers. Aspects of this are long-term contracts,

more information sharing, and a reliance on the design capability of the

first tier suppliers.

Klier draws on several studies to illustrate these points. Eckard (1984)10 is

used to show that the value added  sales ratio in Japan (just-in-time) is 15–

20 per cent, while the North American (Fordist) ratio was 36–48 per cent

10 See page 22 in Klier. Not referred to in Klier's reference list.

 15

before the just-in-time era. The contractual arrangements are explained by

drawing on Asanuma (1989) and Aoki (1988). Eckard (1984) and

Dertouzos et al. (1989) provide the evidence on information efficiency. The

only thing missing is a discussion of why the just-in-time system has

evolved. Is it a good idea no one thought of before (except at Toyota), or

are there underlying drivers such as modern information technology that

explain why this paradigm is winning? Womack, Daniel and Roos (1990)

discuss these issues.

4.1.3 Implication of Changes

Finally, Klier discusses the implications of the paradigm shift on pages 25–

32 with a particular emphasis on the US market. He correctly observes that

based on his earlier discussion, the level of vertical integration should

increase when just-in-time systems are implemented. The reason for this is

that the asset specificity (see Section 4.2, below, on transaction costs)

should increase, and as a consequence it should be more advantageous to

make components and sub-assemblies in-house. Klier then argues that this

is not happening because OEMs and their suppliers are entering into new

contractual forms that make vertical integration unnecessary.

A number of studies support which this argument are effectively

referenced by Klier. Helper (1991), through two surveys comparing 1984

 16

and 1989, shows that suppliers became increasingly likely to provide

information to their customers. The average length of contracts also

increased from 1.2 years in 1984 to 2.3 years in 1989. Furthermore, the

most important purchasing criteria changed from price to quality over the

same time period. Womack, Daniel and Roos (1990) demonstrate how the

number of suppliers is declining and how the US automotive industry is

converging towards a Japanese supplier pattern. Finally, Estall (1985)

studied the locational patterns in the United States and found that

logistical proximity between OEM and supplier is increasing.

This overview of the changes is compelling, but Klier does not explain

why the changes are happening, only that they are happening. In fact,

Mahoney (1989) illustrates that there does not have to be a contradiction

between increased asset specificity and lower vertical integration. Klier

would have made a better argument if he had referenced this work as well

as some of the other research referred to in Mahoney’s thesis.

In summary, though, the overview of the paradigm shift and its

implications is to the point, solidly researched, and well written.

 17

4.2 TRANSACTION COST THEORY

Klier discusses the relevant parts of transaction cost theory in Chapter 2

and the first part of Chapter 3, in total 27 pages covering the underlying

theory and empirical studies.

4.2.1 Theory

This could have been a lengthy discussion because much has been written

on the theory over the last 20 years, and vertical integration problems have

been at the core of this effort. Fortunately, Klier summarises the key

principles of transaction cost theory on pages 33–36 and 55–60 by only

reviewing the cornerstone contributions of Coase (1937) and Williamson

(1975, 1985). He correctly notes that according to their definition of the

theory 1) it is the comparative differences in transaction costs that explain

whether a firm chooses to procure in the market or produce internally

(Coase); and 2) this comparative cost difference is driven by asset

specificity, uncertainty and frequency of transactions (Williamson).

However, there are a number of theoretical objections to this definition

and other definitions exist. Klier should have referred to this and then

made the argument for choosing the particular definition above. An

example of an alternative is Alston and Gillespie (1989), which offers an

 18

extended framework that categorises transaction costs by factors of

production along the production process. The key difference to

Williamson’s framework is that they define the nature of the costs at

various stages in the production process rather than under which

conditions market or internal costs exist. Another example is agency

theory (Jensen and Meckling 1976), which identifies other drivers of

vertical integration such as misalignment of incentives. In fact, Williamson

(1984) makes a similar point within the transaction cost framework. In the

end, Klier would probably have come to the same choice he did since most

empirical work use the definition above, but his choice would have

seemed more informed with an expanded discussion.

4.2.2 Empirical Studies

The emphasis of the literature review is on the ten relevant empirical

studies that address the issue of vertical integration within contexts similar

to what has happened in the automotive industry with the introduction of

just-in-time techniques. Klier also refers to a number of studies which

potentially could have been relevant to his research and explains why this

is not the case. There is at least one important omission though.

Pages 36–54 review the ten studies. Among these ten studies are the

widely quoted ones by Monteverde and Teece (1982), Masten (1984) and
 19

Masten, Meehan and Snyder (1989, 1991). Each of these operationalises the

drivers of transaction costs (asset specificity, uncertainty and frequency)

and tests their explanatory power in a static setting. Two of the studies are

from the automotive industry. Thus, they are highly appropriate to Klier

and the only difference to his research is that he is testing for dynamic

effects.

A few observations are warranted without discussing each of Klier’s ten

reviews. He makes a thorough assessment of Monteverde and Teece (1982)

and notes that while their methodology was sound their sample was

small. Thus, the significance of their test was not strong. Unfortunately,

this comment is not carried forward into the statistical part of the thesis

that relies heavily on Monteverde and Teece’s article.

Klier also reviews Masten, Meehan and Snyder (1991) extensively and

notes that their study not only tested under what conditions a firm

chooses internal or market transactions, but also quantified the monetary

impact of these choices. It could have been worthwhile to discuss why this

approach was not appropriate in Klier’s study, especially since this study

has been cited as the most interesting empirical study of transaction costs

(Joskow 1991).
 20

Finally, a notable omission from the review are the two studies in the

automotive industry by Walker and Weber (1984, 1987) who focus on the

uncertainty aspect of transaction costs. They use a methodology similar to

Monteverde and Teece in their first study and make the important

distinction between demand uncertainty and technological uncertainty.

They find that high demand uncertainty leads to vertical integration while

technological uncertainty has little effect. In their second study they

determine that demand uncertainty only is important when markets are

thin. Thus, uncertainty does not, on balance, appear to be an important

factor. Perhaps this finding should have influenced Klier’s statistical

approach.

In conclusion, as with the review of just-in-time research, Klier has

successfully identified a narrow set of relevant research and extracted the

main findings in a succinct manner. The suggestions for improvement are

not material.
 21

5. ANALYTICAL APPROACH

The analysis section is the weakest part of the thesis. It is sufficient to

support the hypothesis, but it is neither creative nor comprehensive. The

statistical discussion, however, is excellent.

5.1 ANALYTICAL METHOD

Klier needs to demonstrate two things to verify or reject his hypothesis.

First, he needs to show that the preferred governance model has changed

over time towards quasi-integration. Second, he has to demonstrate that

these changes are caused by the introduction of just-in-time techniques.

5.2 DATA COLLECTION

The underlying data appears to be of dubious quality and there is neither

validation of the quality of the data nor a discussion of what would have

constituted a data sample with higher integrity. In fairness, data in the

other related surveys is not always of sufficient quality, so Klier’s sample

is good in relative terms but poor in absolute terms.

 22

5.2.1 Current Data

Klier chose to test his hypothesis by collecting data from the three

domestic US car manufacturers. He collected 89 observations from various

organisational units (there is no mention of response rates) who were

asked to fill out a questionnaire (Klier 1993, 123–132). The questionnaire

only asks for qualitative assessments, usually on a scale from 1 to 5. For

example, the definition of 5 in the assessment of engineering effort is “a lot

of engineering effort”.

The definitions are thus fairly loose and open to a wide range of

interpretation by the people responding to the questionnaire. Perhaps this

was the only way to get information from the car manufacturers, but the

combination of the small sample and the open-ended questionnaire

certainly has reduced the data integrity.

5.2.2 Earlier Studies

Klier also had to draw on data from earlier research to complement his

current industry survey. However, earlier studies have not tested Klier’s

hypothesis explicitly, and thus he had to go back to the source material to

find ways to extract information that helped him test his hypothesis. Both

Monteverde and Teece (1982) and Masten, Meehan and Snyder (1991)
 23

provided such data. Unfortunately, data from the most closely related

study in the automotive industry (Masten, Meehan and Snyder 1989) was

not made available.

Despite the fact that this earlier data was quite different from what he

actually needed, Klier identified the opportunity and collected the

information. However, the dissimilarities make meaningful statistical

comparisons difficult.

5.3 STATISTICAL ANALYSES

While there are limitations in the choice of research methodology, and the

underlying data used is somewhat weak, the statistical analysis is solid.

Klier clearly explains why he chose the methods used, discusses the merits

and problems of each choice made, and applies the statistical methods

rigorously.

Klier’s dependent variables are discrete. He is investigating the changes in

the propensity to vertically integrate, quasi-integrate, or not integrate,

given the introduction of just-in-time techniques. Thus, he correctly uses

qualitative response models which allow for discrete dependent variables,

rather than linear regression models which are intended for continuous

dependent variables (Greene 1993, 635).

 24

5.3.1 Comparison to Previous Studies

Klier needs to compare his results to previous studies (Monteverde and

Teece 1982; Masten, Meehan and Snyder 1989) since he does not have

longitudinal data in his own survey. He does this by identifying the

influence mutual commitment has on governance choice in his own study,

and by comparing the marginal effects (partial derivatives) of the

independent variables in his own study with the previous studies. Klier

uses a binomial probit model in the form of

Probability (vertical integration) = α + β1 · human asset specificity + β2 ·

physical asset specificity + β3 · site specificity + β4 · mutual commitment

+ β5 · frequency + β6–9 · control variables + ε

where vertical integration is 1 if there is vertical integration and 0 if there

is quasi-integration or no integration, the specificity and frequency

variables are according to Williamson’s framework11 and mutual

commitment is the new variable Klier wants to test.

There are a few problems with this approach. First, Klier has to collapse

“quasi-integration” and “no integration” into one state since the earlier

studies did not make the distinction between these two states. This

11 See earlier discussion in Section 4.2.1, above.

 25

reduces his ability to draw inferences from the analysis. He circumvents

this problem by looking at the marginal effects of the independent

variables. Unfortunately, he is only able to calculate the marginal effect of

human asset specificity.

Second, even though he states that the two earlier studies used a binomial

probit method, only Monteverde and Teece did. Masten, Meehan and

Snyder use OLS, two-limit tobit, and logit methods. It is unclear from the

text if Klier has made adjustments for this.

Third, there is confusion about the data source for marginal effects for

Masten, Meehan and Snyder. Klier is using their 1991 study of a naval

construction shipyard,12 not the 1989 study of the automotive industry. It

is unclear why this is the case.

An alternative approach is, in my mind, to work solely with the

Monteverde and Teece data and to expand the data set used when

comparing with Klier. The original article uses the independent variable

specific, which indicates if a part is specific to a single assembler. This

could possibly have been used as an operationalisation of physical asset

specificity. The way to do this practically is to convert Klier’s five-point

12 See page 74 in Klier. Not referred to in Klier's reference list.

 26

scale for physical asset specificity into a dummy variable matching

Monteverde and Teece’s definition. Alternatively, perhaps Monteverde

and Teece’s data can be appended to fit Klier’s more rigorous definition.

5.3.2 Just-in-Time’s Impact on Governance

The second part of Klier’s statistical analysis aims at showing whether the

introduction of just-in-time techniques has changed the relationship

between vertical parties. The choice of models for this part of the research

is not controversial. The three states are clearly ordered and the use of an

unordered probit model would destroy information. Consequently, he

uses two multinomial choice models: ordered probit and sequential probit.

These models allow Klier to use three states in the dependent variable:

vertical integration, quasi-integration, and no integration. The benefit is

that he can thus estimate to what extent just-in-time techniques, with its

dependence on mutual commitment, leads to quasi-integration as

predicted in his hypothesis. The ordered probit model is used under the

assumption that the buyer and supplier decide at one point in time which

governance model to use. The sequential probit model assumes that a first

choice is made between internal and market transaction. Only later, in the

case of market transactions, is a choice made between quasi-integration

and no integration.
 27

Klier uses a similar latent regression as in the binomial probit.13

Probability (vertical integration) = α + β' · x + ε

where vertical integration is 0 for no integration, 1 for quasi-integration,

and 2 for vertical integration, b' is the coefficient vector and x is the

independent variable vector. The key estimation problem is to identify the

threshold value at which no integration becomes quasi-integration.14

One way to improve on Klier’s statistical analysis is to follow Masten,

Meehan and Snyder’s (1991) approach. In their analysis of make-or-buy

decisions at a shipyard, they used a censored regression model. This

allowed them to incorporate cost data for those parts that were made

internally at the shipyard (i.e. vertical integration), while not having to

collect similar data for external parts. The approach makes it possible to

quantify in monetary terms the impact of different vertical relationships.

Clearly, Klier does not have access to such data from the OEMs he

surveyed, and it is unlikely that he can get it. This approach suggests a

way to carry the research on just-in-time‘s impact on vertical relationships

forward. Perhaps Klier should also have made reference to this approach

in his thesis.

13 Klier's definition on page 77 is incorrect. It should include α.

14 The findings section later show that this is not a problem in this analysis.
 28

5.4 MEASUREMENT OF VARIABLES

Klier operationalises his variables in a reasonable way. Human asset

specificity is measured as engineering effort, following the original

definition by Monteverde and Teece (1982, 210). Physical asset specificity

agrees with Masten, Snyder and Meehan’s (1989, 269) definition of

specificity as either due to the car’s style or due to the technical design.

Site specificity is measured as the distance between the supplier’s plant

and the plant of the auto assembler, in line with Spiller (1985).

The mutual commitment variable does not have a precedent in research.

Klier argues that a reasonable way to operationalise the variable is to use

to frequency of delivery between supplier and OEM. The frequency of

delivery is a key characteristic of just-in-time operations (see Helper 1991)

and it qualifies as a proxy for mutual commitment. However, Helper

identifies four other proxies for commitment: the average number of firms

competing to produce a product for a given customer, the average contract

length, supplier costs of switching customers, and criteria used to choose

suppliers. Klier can possibly use one of these instead of frequency of

delivery since they arguably are more important characteristics of just-in-

time operations than his proxy is.

 29

Finally, Klier operationalises the frequency of transaction variable with the

same proxy as above, frequency of delivery. This is not wholly logical

since frequency of transaction in Williamson’s (1985, 60) definition refers

to frequency of contracting. Fortunately, frequency has been shown to be

secondary to asset specificity in importance (Williamson 1985, 52).

In summary, Klier operationalises his variables based on previous research

and theory. There is only limited reason to suggest modifications.

5.5 QUALITY OF STATISTICAL ANALYSIS

The statistical methods are applied rigorously and all key results are

reported in tables. The discussion of the results is thorough and leaves

nothing to ask for. Klier also tests the validity of the results and

documents most of the standard measures of interest such as correlation

coefficients, standard deviations, and significance levels (Klier 1993, 109).

The results are also compared with Monteverde and Teece (1982) and

Masten, Meehan and Snyder (1991) and are found to correlate fairly

closely (Klier 1993, 75, 101).

Klier also discusses problems with his analysis, for example:

...several caveats apply to such a comparison. For example,

different sampling procedures were used across studies...In
 30

addition, each of the studies probably received its

information from a different set of respondents at the
automobile company. (Klier 1993, 76)
 31

6. FINDINGS

Klier wants to test whether the introduction of just-in-time techniques has

changed the nature of vertical relationships in the US automotive industry.

He discusses the findings in Chapter 4B and 5 (pp. 98–119). His overall

conclusion is that the vertical relationships have shifted from a short-term,

arms-length model to a long-term, mutual commitment model. He finds

that his proxy for mutual commitment, frequency of delivery, is

consistently inversely related to the probability of vertical integration, in

line with his hypothesis. Thus, he argues that the strong anecdotal

evidence built over the previous decade has been validated by his

empirical findings. The question is how robust and significant these

findings are. To understand this it is necessary to once again review the

theoretical construct, the operationalisation of the hypothesis, the

statistical significance of the findings, and the quality of the data, to create

a complete picture.

The theoretical construct is highly robust and anchored in previous

research. While the discussion of the theory misses some of the

alternatives, it is still likely that Klier would have picked the same

theoretical foundation in the end. There is a plenitude of research along

similar lines and Klier builds on the work of giants.

 32

When Klier operationalises the theory the robustness decreases

significantly. The use of frequency of delivery as a proxy for commitment

is questionable and there are several alternatives, referred to in Section 5.4,

above. Commitment is the most important variable in Klier’s hypothesis

and one can argue that frequency of delivery is nothing else than

frequency of delivery. The other proxies are more solid, but also less

important.

As discussed in Chapter 5, above, Klier’s analysis is somewhat, but not

highly, statistically significant. The binomial probit analysis has a high

level of significance for mutual commitment and human asset specificity

(p. 99). However, this analysis is not optimal since it does not explicitly

take into account quasi-integration. Moreover, it is only by studying the

partial derivatives of the variables that comparisons can be made over

time. The ordered probit and sequential probit analyses are less significant

but do not refute the hypothesis.

Finally, the quality of the underlying data is moderate. The survey is

qualitative and leaves room for interpretation, the data is difficult to

compare to earlier studies, and the sample size is not wholly adequate.

On balance, Klier’s findings are not robust but they do add insights on the

margin. The results are in line with earlier research and contribute to the
 33

overall picture of transaction cost economics ability to shed light on

vertical integration issues. But as always, more research will be required.

One avenue for this is to replicate Klier’s study but with a different set of

dependent variables. Another is to do a more in depth study of one

automotive company, using inside data. This will allow a quantification of

the economic impact, and the granularity of the analysis can be improved.

Perhaps this particular aspect of the vertical integration problem has been

exhausted, though.
 34

7. THESIS LAYOUT

The thesis covers 119 pages (excluding two short appendices) and roughly

half of it is dedicated to the literature survey, the rest to building and

testing the hypothesis. This composition is appropriate since it gives

sufficient attention to all parts of the research.

The mechanics applied are excellent. Klier uses good academic grammar,

there are few—if any—spelling errors, and quotes and footnotes are

correct and relevant. The reference list is mostly accurate, although the

source for Coase (1937) is wrong (should be Economica, not

Econometrica), and both Eckard (1984) and Masten, Meehan and Snyder

(1991) is referred to in the text but is missing in the references. A minor

point is that the graphical layout is unappealing and it is difficult to follow

chapters, headings, etc., since one typeface and font size is uniformly

applied.

The main objection to the layout is the structure of Chapters 2 and 3.

Chapter 2 covers most of the review of transaction cost theory. However,

this review continues in the beginning of Chapter 3 and Klier then

continues to describe his hypothesis at the end of Chapter 3. This creates

both repetitiveness and it somewhat hides the hypothesis.

 35

In addition, the description of the choice of statistical methods is either

overly detailed (if the aim is to give a summary description) or too short (if

Klier hopes to explain the intricacies of the choices he makes).

On balance, though, the power and clarity of the language makes the

thesis accessible and interesting.

 36

8. CONCLUSION

Klier has identified an important problem, articulated a strong hypothesis,

and offers compelling, but not conclusive, evidence to support the

hypothesis. The whole thesis is targeted at this task and there is little

excess discussion. The findings are in line with earlier findings although

the weak underlying data makes it difficult to see far-reaching

implications.

Overall, Klier has furthered our knowledge of the implications of just-in-

time techniques and contributed to making transaction cost theory more

operational.
 37

REFERENCES

Alston, L. J. and W. Gillespie. 1989. Resource Coordination and

Transaction Costs: A Framework for Analyzing the Firm/Market
Boundary. Journal of Economic Behavior and Organization 11 (2):
191–212.

Aoki, M. 1988. Information, Incentives, and Bargaining in the Japanese

Economy. Cambridge: Cambridge University Press.

Asanuma, B. 1989. Manufacturer–Supplier Relationships in Japan and the

Concept of Relation-Specific Skills. Journal of the Japanese and
International Economies 3 (1): 1–30.

Balakrishnan, S. and B. Wernerfelt. 1986. Technical Change, Competition

and Vertical Integration. Strategic Management Journal 7 (4): 347–359.

Caves, R. and R. E. Bradburd. 1988. The Empirical Determinants of

Vertical Integration. Journal of Economic Behavior and Organization
9 (3): 265–279.

Coase, R. H. 1937. The Nature of the Firm. Economica n.s., 4 (16): 386–405.

Cohen, S. S. and J. Zysman. 1987. Manufacturing Matters: The Myth of the

Post-Industrial Economy. New York: Basic Books.

Dertouzos, M. L., R. K. Lester, R. M. Solow and the MIT Commission on

Industrial Productivity. 1989. Made in America: Regaining the
Productive Edge. Cambridge: MIT Press.

Eckard, E. W. 1984. Alternative Vertical Structures: The Case of the

Japanese Auto Industry. Business Economics 19 (October): 57–61.

Estall, R. C. 1985. Stock Control in Manufacturing: The Just-in-Time

System and Its Locational Implications. Area 17 (2): 129–132.

Greene, W. H. 1993. Econometric Analysis. 2d ed. Englewood Cliffs, N.J.:

Prentice Hall.

Helper, S. 1991. How Much Has Really Changed between U.S.

Automakers and Their Suppliers? Sloan Management Review 32
(Summer): 15–28.

Hirschman, A. O. 1970. Exit, Voice, and Loyalty: Responses to Decline in

Firms, Organizations, and States. Cambridge: Harvard University
Press.
 38

Jensen, M. C. and W. H. Meckling. 1976. Theory of the Firm: Managerial

Behavior, Agency Costs and Ownership Structure. Journal of
Financial Economics 3 (4): 305–360.

Joskow, P. L. 1991. The Role of Transaction Cost Economics in Antitrust

and Public Utility Regulatory Policies. Journal of Law, Economics, and
Organization 7 (Special issue): 53–83.

Klier, T. H. 1993. Transaction Cost Theory and Just-in-Time

Manufacturing: A New Look at Vertical Integration in the United
States Automobile Market. PhD thesis, Department of Economics,
Michigan State University.

Mahoney, J. T. 1989. Organizational Rationalization and Innovation:

Vertical Integration and Multidivisional Organization. PhD thesis,
University of Pennsylvania.

———. 1992. The Choice of Organizational Form: Vertical Financial

Ownership versus Other Methods of Vertical Integration. Strategic
Management Journal 13 (8): 559–584.

Masten, S. E. 1984. The Organization of Production: Evidence from the

Aerospace Industry. Journal of Law and Economics 27 (October):
403–417.

Masten, S. E., J. W. Meehan, Jr. and E. A. Snyder. 1989. Vertical Integration

in the US Auto Industry: A Note on the Influence of Transaction
Specific Assets. Journal of Economic Behavior and Organization 12 (2):
265–273.

———. 1991. The Costs of Organization. Journal of Law, Economics and

Organization 7 (1): 1–25.

MIT Commission on Industrial Productivity. 1989. The U.S. Automobile

Industry in an Era of International Competition: Performance and
Prospects. In The Working Papers of the MIT Commission on Industrial
Productivity. Cambridge: MIT Press.

Monteverde, K. and D. J. Teece. 1982. Supplier Switching Costs and

Vertical Integration in the Automobile Industry. Bell Journal of
Economics 13 (1): 206–213.

Scherrer, C. 1991. Governance of the Automobile Industry: The

Transformation of Labor and Supplier Relations. In Governance of
the American Economy, edited by J. L. Campbell, J. R. Hollingsworth
and L. N. Lindberg. Cambridge: Cambridge University Press.
 39

Spiller, P. T. 1985. On Vertical Mergers. Journal of Law, Economics, and

Organization 1 (2): 285–312.

Walker, G. and D. Weber. 1984. A Transaction Cost Approach to Make-or-

Buy Decisions. Administrative Science Quarterly 29 (3): 373–391.

———. 1987. Supplier Competition, Uncertainty and Make-or-Buy

Decisions. Academy of Management Journal 30 (3): 589–596.

Williamson, O. E. 1975. Markets and Hierarchies: Analysis and Antitrust

Implications. New York: Free Press.

———. 1984. The Incentive Limits of Firms: A Comparative Institutional

Assessment of Bureaucracy. Weltwirtschaftliches Archiv 120 (4):
736–763.

———. 1985. The Economic Institutions of Capitalism. New York: Free Press.

Womack, J. P., D. T. Jones and D. Roos. 1990. The Machine That Changed the
World. New York: Rawson.